J. Dairy Sci.
101:1–9
https://doi.org/10.3168/jds.2018-15045
© American Dairy Science Association®, 2018.
Identification of objectionable flavors in purported
spontaneous oxidized flavor bovine milk
David M. Potts* and Devin G. Peterson†1
*Department of Food Science and Nutrition, University of Minnesota, St. Paul 55108
†Department of Food Science and Technology, The Ohio State University, Columbus 43210
ABSTRACT
Spontaneous oxidized flavor (SOF) has been reported
over the past 5 decades as a sporadic objectionable
flavor problem in bovine milk. Parameters previously
reported to influence SOF development in milk have
been contradictory, limiting the ability to monitor and
develop mitigation strategies. The current paper inves-
tigates the causative compounds associated with milk
identified as SOF milk in the Midwest dairy region of
the United States. Based on GC/MS-olfactometry anal-
ysis, endo-borneol, 2-methylisoborneol, and α-terpineol
were identified as the off-flavor compounds. Sensory
recombination studies further confirmed the sensory
contribution of these compounds to the noted off-flavor
attributes in the original milk, which were described
as “green,” “musty,” and “unclean.” These compounds
are known microbial-derived flavor taints, indicating
oxidation was not the origin of the objectionable flavor
in the milk. This noted misclassification of the milk as
SOF indicates the challenge of defining flavor defects
without the identification of the active compounds.
Key words: milk, off-flavor, oxidation, spontaneous
oxidized flavor, microbial taint
INTRODUCTION
Spontaneous oxidized flavor (SOF) in bovine milk
has been identified in recent decades as a sporadic
off-flavor problem affecting milk both before and after
pasteurization. It is primarily characterized by “card-
board,” “fishy,” or “oxidized” flavors, and it is often
suggested as capable of contaminating the flavor of
otherwise acceptable milk during bulk tank mixing
before pasteurization. The off-flavor was first identified
in northern Europe, and during the last decade it has
been recognized as a growing problem in the Midwest-
Received May 11, 2018.
Accepted August 1, 2018.
1
Corresponding author: peterson892@​osu​.edu
1
ern United States. It has been frequently described as a
seasonal off-flavor taint, appearing in the late autumn
months and disappearing in early spring (Juhlin et al.,
2010a). Because the associated objectionable flavors
are reminiscent of lipid oxidation and affect the flavor
of clean milk, a common assumption is an oxidative
radical mechanism is responsible, hence the name SOF.
Researchers have frequently attempted to correlate
SOF sensory defects with changes in known causative
factors associated with oxidative reaction pathways.
The most common examples include changes in bovine
diet, such as the presence of specific PUFA, as well
as other factors that affect instigation or inhibition of
SOF through various oxidizers, antioxidants, catalysts,
and so on (Dunkley et al., 1963; Bruhn et al., 1976;
Juhlin et al., 2010a).
Several studies have suggested that the presence of
particular micronutrients such as copper may catalyze
lipid oxidation products in SOF milk. Juhlin et al.
(2010a) claimed that the problem might be partially
genetic, especially in the variability of copper levels
during lactation of individual cows. In a later study, the
group also asserted that a significant correlation was
present between PUFA, copper levels, and SOF forma-
tion, although they noted that precisely defining the
levels of substrates and pro- and antioxidants directly
responsible was exceedingly complex and required fur-
ther study (Juhlin et al., 2010b). Timmons et al. (2001)
further described a link between increased SOF with
higher levels of PUFA combined with a higher presence
of copper within milk.
Others have attributed SOF formation exclusively to
lipid-antioxidant levels and interactions. Clausen et al.
(2010) found that higher concentrations of initial lipid
peroxidation products, in combination with reduced
levels of lower molecular weight antioxidants in the
initiation stages of lipid oxidation, could accelerate oxi-
dized flavor defects. In contrast, Granelli et al. (1998)
investigated levels of PUFA, tocopherol, and β-carotene
levels, concluding that their control herds showed lower
antioxidant:​PUFA ratios than experimental herds, yet
did not develop any characteristic SOF flavor.
2 POTTS AND PETERSON
A limitation of prior studies focused on SOF instiga-
tion and inhibition has been the lack of research focused
on the identification of active compounds. Knowledge of
the compounds responsible for the flavor defects in SOF
milk would lend further insight regarding the potential
source and enabling the development of effective miti-
gation strategies. Furthermore, it also provides a basis
to analytically monitor for product defects. The overall
objective of this study was to identify the compounds
responsible for the flavor defects in the milk sample ob-
tained from the Midwestern US region flagged as SOF
by Midwest Dairy Foods Research Center.
MATERIALS AND METHODS
Milk Samples
Nonflavor defective (control) and off-flavored milk
samples labeled as SOF were obtained through the
Midwest Dairy Foods Research Center (St. Paul, MN).
The SOF labeled milk was obtained from a farm in ru-
ral Nebraska flagged by producers for several consecu-
tive years as having a noticeable, cardboard/oxidized
SOF-type flavor taint arising from Jersey and Holstein
cows between late autumn and early spring. Both the
control and SOF labeled raw whole milk samples were
shipped in chilled containers to the University of Min-
nesota for analysis. Upon arrival, samples were evalu-
ated (and expectorated) by 3 experienced trained dairy
judges at the University of Minnesota (Department of
Food Science and Nutrition) to evaluate overall flavor
quality. The control milk was noted as having typical
flavors associated with high-quality whole fat milk, and
the SOF labeled milk was noted as having “unclean” or
poor flavor quality and aftertaste, as well as a metal-
lic sensation. Milk samples were HTST pasteurized at
the University of Minnesota Joseph Warthesen Food
Processing Center (St. Paul, MN) using a HTST/UHT
Direct Steam Injection & Indirect Tubular Processing
System (MicroThermics, Raleigh, NC). Processing
conditions included sanitation using a chlorine solution
(100 ppm) followed by a clean water wash at 248°F
(120°C) for 1 h. Control and SOF labeled milk samples
were individually homogenized and processed to a final
temperature of 165°F (74°C) for 20 s to ensure proper
pasteurization. Previously sterilized glass bottles were
filled with respective samples and cooled immediately
to 60°F (16°C). Next, samples were tasted by 3 expe-
rienced dairy judges to evaluate flavor quality of the
control and SOF labeled samples postpasteurization.
The judges identified the control milk as “clean,” or
containing no objectionable volatile notes; the SOF
labeled sample, however, was identified as retaining the
“unclean” flavor profile with a metallic and objection-
Journal of Dairy Science Vol. 101 No. 12, 2018
able aftertaste. The pasteurized samples were either
immediately prepared for analysis (d 0 samples) or
placed under refrigerated storage (4–5°C) for 2 wk be-
fore subsequent extraction and analysis (d 14 samples).
A standard plate count methodology was conducted
during the 2-wk period using a Plate Count Agar
(Neogen Corporation, Lansing, MI) to observe overall
colony-forming units per milliliter and ensure adequate
pasteurization of samples. For both control and SOF
labeled milks, plate counts were initially calculated as
too few to count and 2,400 cfu/mL, respectively; both
fell within the <20,000 cfu/mL limits for grade A pas-
teurized milk as dictated by the United States Pasteur-
ized Milk Ordinance. After 14 d of storage, plate counts
were calculated as 1,800 and 49,000 cfu/mL for control
and off-flavored milks, respectively.
Chemicals
Authentic standards used for quantification in this
study included E-2-decenal, 2-heptanone, nonanal,
2-nonanone, 2-methylisoborneol, 2-methyl-3-hepta-
none, α-terpineol, E,E-2,4-decadienal, 2-methylvale-
ric acid, and endoborneol (Sigma-Aldrich, St. Louis,
MO). Diethyl ether was obtained from Fisher Scientific
(Somerville, NJ).
Extraction Procedure
A modified version of the methodologies reported by
Karagül-Yüceer et al. (2001) and Colahan-Sederstrom
and Peterson (2005) were incorporated for the volatile
extraction of control and SOF labeled milks. Samples
were extracted in duplicate after pasteurization, both
fresh (d 0) and after 2 wk (d 14) to observe storage
effects on any off-flavors present, particularly because
SOF has historically been reported to exacerbate over
time. One liter of redistilled diethyl ether was spiked
with internal standard solutions: 3 µL of a 2-methyl-
3-heptanone internal standard solution (50 µL/5 mL
of diethyl ether) for the neutral/basic volatile extract,
and 3 µL of a 2-methylvaleric acid internal standard
solution (33 µL/5 mL of diethyl ether) for the acidic
volatile extract. One liter of each respective milk sam-
ple was measured and divided into 5 individual Teflon-
lined centrifuge bottles with Tefzel closures (Nalgene,
Rochester, NY). Twenty grams of sodium chloride was
added in each bottle, followed by addition of 40 mL
of the redistilled diethyl ether with the internal stan-
dards (Fisher Scientific, Fair Lawn, NJ). The milk in
each bottle was individually extracted 5 times (40 mL/
bottle, 5 × 200 mL extractions). After solvent addi-
tion, each bottle was gently overturned several times,
 OBJECTIONABLE FLAVORS IN BOVINE MILK
followed by mild stirring on an orbital shaking table
(model 3540, Labline Instruments Inc., Melrose Park,
IL) at 30% maximum speed for 20 min, and gently
overturned several times again to avoid emulsion for-
mation. The samples were centrifuged for 20 min at
4°C at 3,000 × g (Allegra X-22R Centrifuge, Beckman
Coulter, Brea, CA). After centrifugation, the solvent
was carefully removed and pooled, dried over anhy-
drous sodium sulfate, filtered, and stored overnight in
a −20°C freezer. Samples were then concentrated to
approximately 100 mL using a fractional distillation
column. The volatile fraction was then isolated using a
solvent-assisted flavor evaporation system as described
by Engel et al. (1999). The resulting volatile fractions
were further concentrated to approximately 20 mL,
after which they were additionally fractionated into
neutral/basic and acidic fractions. The ether samples
were washed with 1 M sodium bicarbonate (3 × 50
mL) and a saturated solution of sodium chloride in
water (2 × 55 mL) using a separatory funnel. The or-
ganic phase containing the neutral/basic volatiles was
isolated and dried over anhydrous sodium sulfate and
further distilled to a volume of approximately 1 mL.
The aqueous fractions were combined and reacidified
to a pH of 1.5 with 18% vol/vol of 12 M hydrochloric
acid and extracted with repurified diethyl ether (3 × 50
mL) to isolate the acidic volatile fraction. The fraction
was dried over anhydrous sodium sulfate and distilled
to approximately 1 mL. Extracts were stored at −80°C
before analysis.
Gas Chromatography-Mass Spectrometry
Flavor extracts were analyzed with an Agilent 7890A
GC (Agilent Technologies, Santa Clara, CA) equipped
with a Leco Pegasus 4D GCXGC 4D TOF-MS system,
as well as a Agilent Technologies 6890 GC equipped
with a Agilent Technologies 5973 MSD. Two μL of
each sample was injected under splitless mode; inlet
temperature was set at 250°C, and helium was used
as a carrier gas with total flow rate of 1.0 mL/min.
Compounds were analyzed on 2 different GC capillary
columns with different chemistries: a DB-WAX (Agi-
lent, 60 m × 0.25 mm × 0.25 μm) and HP-5 (Agilent,
60 m × 0.25 mm × 0.25 μm). The temperature profile
used for the DB-WAX column was 40°C for 2 min, then
gradually increased to 220°C at 3°C/min and was held
for 3 min; for the HP-5 the oven temperature profile
started at 40°C and was held for 2 min, then gradually
increased to 250°C at 3°C/min and was held for 3 min.
Mass spectral data were obtained in scan mode with a
range of 30 to 350 amu. Confirmation of compounds
was performed using authentic standards coupled with
mass spectral data.
3
Gas Chromatography-Mass
Spectrometry/Olfactometry
Flavor extracts were analyzed using an Agilent Tech-
nologies 6890A GC coupled with an Agilent Technolo-
gies 5973 MS detector and equipped with an olfactom-
etry port (Gerstel, Baltimore, MD). Chromatographic
separation was performed on a DB-WAX or an HP-5
column (60 m × 0.25 mm × 0.25 µm). One microliter
of each sample was injected under splitless mode with
the inlet temperature set at 250°C. The temperature
profile used for the DB-WAX column was 40°C for 2
min, then gradually increased to 230°C at 5°C/min and
held for 3 min. For the HP-5 column, the temperature
profile was 40°C for 2 min, then gradually increased to
250°C at 3°C/min and held for 3 min. After separation,
the effluent of each column was split 1:1 between an MS
and olfactory sniffing port. Purified air was bubbled
through distilled water and purged at the end of the
sniffing port at a rate of 10 mL/min. Three panelists
experienced in GC-MS/olfactometry analysis protocols
identified aromas as they eluted from the heated olfac-
tory sniffing arm (250°C). During olfactory runs, panel-
ists recorded the elution time of the compound, per-
ceived odor descriptor, and intensity on a scale from 1
to 5, based on the OSME method for GC-olfactometry
(McDaniel et al., 1990). Intensity ratings were assigned
numerical values as follows: 1 = very weak; 2 = weak; 3
= moderate; 4 = strong; and 5 = very strong.
Data from each panelist were compiled and values
were averaged among all 3 individuals. Compounds
were positively identified using mass spectral data,
odor descriptors, and Kovats retention indices in com-
parison with authentic standards. Compounds were
labeled as “tentatively identified” if mass spectra were
not obtained.
Quantitation of Selected Off-Flavor Compounds
The selected compounds endo-borneol, 2-methyliso-
borneol, α-terpineol, 1-nonanal, 2-heptanone, 2-nona-
none, E-2-decenal, and E,E-2,4-decadienal were quan-
tified in the SOF labeled and control milk samples.
Analyses were performed on an Agilent Technologies
6890A GC coupled with an Agilent Technologies 5973
MS detector using either scan mode (30–350 amu) or
selective ion monitoring mode for maximum sensitivity
on a HP-5 column. One microliter of each sample was
injected under splitless mode with the inlet temperature
set at 250°C. The temperature profile used was 40°C for
2 min, then gradually increased to 250°C at 7°C/min
and held for 3 min. Helium was used as a carrier gas
and set at constant flow at 1.0 mL/min. A combination
of internal and external standard protocols was used.
Journal of Dairy Science Vol. 101 No. 12, 2018
4 POTTS AND PETERSON
Quantitative analysis was completed using comparison
of peak areas and the calculation of retention factors
using the internal standards described above. Percent
recovery studies were performed in triplicate and cal-
culated for each of the individual compounds of inter-
est, and the final calculated odorant concentrations
were comparable to the concentrations obtained from
independent 5-point calibration curves for each com-
pound (R2 > 0.97 and good linearity observed for all
compounds). For selective ion monitoring analyses, ions
characteristic to the odorants of interest were selected
and are reported in Table 1. Statistical significance was
verified using the Student’s t-test. All analyses were
performed with Statistix 10 (Analytical Software, Tal-
lahassee, FL). Statistical significance was established at
P < 0.05 for all compounds.
Sensory Recombination Study
Approval for the sensory evaluation protocol was
granted by the Ethics Committee, University of Minne-
sota (IRB #1510E79045). An 11-member sensory panel
was convened, consisting of individuals familiar with
sensory evaluation protocols for various food products.
Panelists participated in a degree of difference sensory
test. They were provided with 30 mL of commercially
available whole milk as a reference. Panelists were
served milk samples randomly assigned with 3-digit
codes. For each test, panelists were presented with
3 different milk samples: control whole milk sample
(same as the reference milk) and 2 recombinant SOF
labeled milk model samples (control milk with added
compounds) that consisted of the objectionable flavors
(identified by GC-MS/olfactometry analysis) at con-
centrations at either d 0 or 14. The control milk was
defined as a high quality, commercially available whole
bovine milk. Panelists were asked to taste the reference
milk, followed by each of the 3 randomly coded samples
from left to right and evaluate them for degree of over-
all flavor difference from the reference (not appearance
or texture). Panelists were instructed to rate the milks
on a 5-point scale, ranging from 0 = no difference, 1 =
very slight difference, 2 = slight difference, 3 = moder-
ate difference, 4 = large difference, or 5 = extremely
large difference. Panelists were additionally asked to
describe the perceived flavor differences noticed in the
samples. Data were analyzed using a 2-sided Dunnett’s
multiple comparisons test, using the statistical software
package Statistix 10 (Analytical Software).
RESULTS AND DISCUSSION
An objectionable flavored milk sample was obtained
from a farm in the Midwestern United States for
Journal of Dairy Science Vol. 101 No. 12, 2018
Table 1. Select aroma compounds quantified for concentration
comparison in control and off-flavored milk samples1
Molecular Monitored ions
Compound name weight (SIM)
2-Methyl-3-heptanone (Istd) 128 128
Nonanal 142 98, 142
Endo-borneol 154 95, 154
2-Methylisoborneol 168 95, 107, 168
α-Terpineol 154 93, 154
E-2-Decenal 154 97, 154
E,E-2,4-Decadienal 152 81, 152
1
Corresponding values for average compound molecular weights and
ions monitored during GC-MS selective ion monitoring (SIM)-mode
analysis for respective peak areas.
analysis. The Midwest Dairy Foods Research Center
designated the milk as SOF based on the documented
seasonality (over 2 consecutive years) and noted flavor
descriptors (“oxidized,” “cardboard,” “fishy,” and so
on; Shipe et al., 1978; Nicholson, 1993; Juhlin et al.,
2010b). Thus, elevated levels of lipid oxidation species
were expected.
The aroma profile of the pasteurized SOF milk and
a control milk (no apparent defects) after 0 and 14
d of storage were characterized by GC-MS/olfactom-
etry analysis. Many of the compounds identified were
common to both samples and are typical of fresh, high
quality milk, including fatty acids, aldehydes, ketones,
and lactones (shown in Tables 2–5; Villeneuve et al.,
2013). The acidic fraction volatile compounds identified
in the SOF labeled milk and control samples at d 0 and
14 are shown in Tables 2 and 3, respectively. The odor
intensities and peak areas (data not shown) revealed
only negligible differences in the profiles of the known
hydrolytic lipid oxidation products between the samples
and during storage. The acids identified were typical to
milk and associated with “fatty” or “cheese-like” aroma
descriptors. Therefore, the free fatty acid compounds
were not considered as the source of the objectionable
flavor and were not further evaluated.
The odor-active compounds detected in the neutral/
basic fraction of the SOF labeled and control milk
samples at d 0 and 14 are shown in Tables 4 and 5,
respectively. Typical in milk, lipid peroxidation com-
pounds were identified with descriptors such as green
and fatty, as well as hydrolytic thermally derived meth-
yl ketone products. In general, many of the compounds
identified in the SOF sample were also present in the
control milk; however, several key differences were also
observed. As a first part of the analysis, differences in
the noted lipid oxidation products were investigated.
Two criteria were established to select lipid oxidation
products for quantification: (1) observed >30% peak
area (total ion current-MS) in SOF versus control or
(2) observed an increase of >30% in peak area (total
 OBJECTIONABLE FLAVORS IN BOVINE MILK 5
Table 2. Aroma compounds identified in the acidic fraction of control milk1

Average OSME intensity

Identified compound   DB-WAX2   HP-52   Odor descriptor d0 d 14

Acetic acid 1490 594 Fat, milk, cheese 2.3 2.2
Isobutyric acid 1556 1195 Fat, cheese, oil 2.8 2.5
Butyric acid 1597 846 Rancid oil 3.2 2.2
Pentanoic acid 1722 976 Cheese 2.3 2.2
Hexanoic acid 1803 1022 Soap, sweat 2.3 2.7
Heptanoic acid 1979 1103 Sweat, fat 2.7 2.8
Octanoic acid 2061 — Fatty 2.3 2.5
Decanoic acid 2296 1401 Rancid, fat 2.5 2.0
Benzoic acid 2434 — Fat, rancid 1.8 2.0
Dodecanoic acid 2483 — Fat 1.8 2.5
1
Identified on 2 column chemistries with corresponding odor descriptors and OSME odor intensity values from d 0 and 14 post-HTST pasteuri-
zation, averaged among 3 panelists.
2
DB-WAX and HP-5: Agilent Technologies, Santa Clara, California.

ion current-MS) over the 14-d period of storage (data
not shown) in the SOF labeled sample. Accordingly, 5
compounds were selected and are shown in Figure 1.
Nonanal and E,E-2,4-decadienal were only observed in
the SOF milk, at concentrations of 9.1 to 11and 0.41 to
0.91 μg/L, respectively. These values were above their
published odor threshold concentrations (1 μg/L and
0.07–0.2 μg/L, respectively; Buttery et al., 1988; Plotto
et al., 2004). However, decadienal decreased signifi-
cantly over the 14-d storage period, whereas no change
in concentration was noted for nonanal. The other 3
compounds, 2-heptanone, 2-nonanone, and E-2-decenal
were not detected in the d 0 samples but were present
after 14 d. The concentrations of both 2-heptanone and
2-nonanone (1.9 and 1.5 µg/L, respectively) were below
the lowest published literature odor threshold values in
water (140 and 5 μg/L, respectively; Teranishi et al.,
1974). The concentrations of E-2-decenal (0.69 µg/L)
was slightly above its published aqueous odor threshold
of 0.3 μg/L (Buttery et al., 1988). However, considering
whole milk contains 3 to 4% fat, the threshold values of
detection would be expected to increase (higher concen-
tration) for these lipophilic compounds in comparison
to a water base. The lack of unique lipid oxidation
compounds in the SOF labeled milk above the detec-
tion threshold, as well as no observed general increase
in lipid oxidation compounds during storage, suggested
the traditional SOF hypothesis was not contributing to
the objectionable flavor defect.
Further review of the odor-active compounds detect-
ed in the neutral/basic fraction revealed the SOF la-
beled milk had a distinct series of terpenoid compounds
with characteristic “green” and “musty” descriptors,
that included endo-borneol, 2-methylisoborneol, and
α-terpineol (Table 5). These compounds were not de-
tected in the control milk, suggesting possible involve-
ment in the perceived objectionable flavor reported in
the SOF labeled milk. These terpenoid compounds are
known flavor taints that originate from bacterial and
fungal sources. For example, 2-methylisoborneol has
been reported to bioaccumulate in catfish, resulting
in sensory defects that originated from cyanobacteria
(blue-green algae) in aqueous environments (Schrader
and Summerfelt, 2010). Siegmund and Pöllinger-Zierler

Table 3. Aroma compounds identified in the acidic fraction of spontaneous oxidized flavor labeled milk1

Average OSME intensity

Identified compound   DB-WAX2   HP-52   Odor descriptor d0 d 14

Acetic acid 1436 689 Sour 2.3 3.2
Butyric acid 1623 847 Fat, rancid, sweat 3.0 3.3
Hexanoic acid 1835 1046 Cheese, fat, grease 2.5 2.5
Heptanoic acid 1941 1107 Humid, sweaty, metal 2.8 2.3
Octanoic acid 2060 — Fat 2.5 1.5
Nonanoic acid 2185 — Cheese, fat, grease 1.5 1.5
Decanoic acid 2270 1400 Fat 1.8 2.0
Benzoic acid 2413 — Fat, creamy 2.3 1.5
1
Identified on 2 column chemistries with corresponding odor descriptors and OSME odor intensity values from d 0 and 14 post-HTST pasteuri-
zation, averaged among 3 panelists.
2
DB-WAX and HP-5: Agilent Technologies, Santa Clara, California.

Journal of Dairy Science Vol. 101 No. 12, 2018

6 POTTS AND PETERSON

Table 4. Aroma compounds identified in the neutral/basic fraction of control milk1

Average OSME intensity

Identified compound   DB-WAX2   HP-52   Odor descriptor d0 d 14

Acetic acid 1450 607 Fat, cheese, pungent — 2.8
Pentanal 902 717 Pungent, fat — 1.7
Isobutyl acetate — 775 Wheat, almond 2.0 1.7
Hexanal 1084 804 Wheat, dough, grain — 2.2
1-Nonen-3-one 1239 836 Fat, milkfat 2.3 3.0
Hexanol — 879 Cabbage, onion 1.7 —
Mercaptomethylbutanol — 974 Brothy, bouillon — 3.0
1-Octen-3-one 1318 984 Mushroom, brothy, roasted 2.3 2.8
Pentylfuran 1245 999 Earthy, green, rubbery 2.3 2.5
Benzyl alcohol — 1034 Roasted, meaty 2.3 1.3
1-Octanol 1547 1061 Sulfury, pungent, barn 2.2 1.5
2-Nonanone 1390 1091 Milk, fat 2.2 3.0
Nonanal 1400 1106 Milk, fat, dairy 2.2 3.3
Menthol 1646 1180 Green — 2.3
Decanal 1450 1188 Green, paper, rancid 2.5 2.5
Z-4-decenal 1522 1195 Rancid oil 2.0 2.3
Carveol — 1220 Musty, rancid, papery 3.0 2.5
γ-Octalactone 1869 1260 Oily, lactone — 2.7
δ-Octalactone 1921 1280 Peach, dairy 2.5 2.7
Thymol 2242 1293 Herbal 2.3 2.8
Geranyl acetone — 1444 Green, musty 2.5 2.8
δ-Decalactone 2235 1453 Sweet, vanilla 3.2 2.7
γ-Decalactone 2135 1483 Lactone, peach 2.8 2.7
Methyl laurate 1786 1511 Fatty 3.7 2.8
1
Identified on 2 column chemistries with corresponding odor descriptors and OSME odor intensity values from d 0 and 14 post-HTST pasteuri-
zation, averaged among 3 panelists.
2
DB-WAX and HP-5: Agilent Technologies, Santa Clara, California.

(2006) identified a wide range of “musty” compounds
present in apple juice both before and after pasteuriza-
tion including endo-borneol, 2-methylisoborneol, and
α-terpineol. In this study, populations of various ther-
moacidophilic bacteria (Alicyclobacillus acidoterrestris,
Actinomycetes, and Streptomyces spp.) were found to
quantitatively increase over time on the surfaces of
apples, which corresponded with increasing sensory
defects in apple juice produced from the contaminated
fruit. In general, terpenoid compounds have been noted
in literature as having exceedingly low odor thresholds
(endo-borneol at 140 μg/L in water, 2-methylisoborneol
at 0.002–0.1 μg/L in water, and α-terpineol at 4.6 μg/L
in water), as well as notable “green” and “musty” odors.
It is also worth noting that they reported the odor
thresholds of these identified terpenoid compounds in
apple juice were perceived at lower concentrations than
previously published odor thresholds in water (Sieg-
mund and Pöllinger-Zierler, 2006).
The average concentrations of the terpenoid com-
pounds in the objectionable milk samples after d 0 and
14 are shown in Figure 2; again, these compounds were
not detected in the control milk. Over the 14-d storage
period, endo-borneol and α-terpineol decreased sig-
nificantly (P < 0.05). Endo-borneol decreased approxi-
mately 5-fold from 0.95 to 0.20 µg/L and α-terpineol
decreased about 26%, from 0.84 to 0.62 µg/L, whereas
Journal of Dairy Science Vol. 101 No. 12, 2018
2-methylisoborneol increased significantly (P < 0.05)
from 0.0037 to 0.017 µg/L, approximately a 4-fold
increase. The concentration of 2-methylisoborneol was
well above its published odor threshold of 0.002 μg/L
in water. The survival of microbial strains in pasteur-
ized milk may have resulted in the noted increase of
2-methylisoborneol during storage (Siegmund and
Pöllinger-Zierler, 2006). The decrease in concentration
of endo-borneol and α-terpineol indicated that these
products were not stable over time in the milk samples.
It is uncertain whether the loss of endo-borneol and
α-terpineol in the milk samples was directly related
to the increase of 2-methylisoborneol over time, either
through microbial metabolism or other conversion
pathways. Further study would be necessary to eluci-
date these mechanisms.
As a first step to screen the effect of the noted
changes in lipid oxidation and terpenoid compounds
on the observed objectionable flavor attributes of the
SOF-labeled milk, a preliminary sensory recombina-
tion study was conducted with 3 experienced dairy
judges. Nonflavor defective milk (control) was dosed
with either the lipid oxidation products (2-heptanone,
2-nonanone, E-2-decenal, E,E-2,4-decadienal, nonanal)
or the terpenoid compounds (endo-borneol, 2-methyl-
isoborneol, α-terpineol), or a combination of all com-
pounds, at equivalent concentrations of the d 0 and
 OBJECTIONABLE FLAVORS IN BOVINE MILK 7

14 SOF-labeled samples. The recombination sample
with the lipid oxidation compounds was not found to
be objectionable by any of the judges, and thus the
compounds were excluded from further sensory evalu-
ation. However, evaluation of the recombination milk
sample with the terpenoid compounds were perceived
as unclean with distinctly unpleasant aftertastes and
matched the initial sensory screening of the purported
SOF milk samples at the time of reception from the
dairy farm. Finally, the recombination sample contain-
ing all selected volatile compounds was described in a
similar manner as the recombination sample with only
terpenoid compounds. Therefore, the terpenoid com-
pounds were selected for further sensory evaluation.
A degree of difference test was employed to deter-
mine the magnitude of sensory change and assess the
relevance of the identified terpenoid compounds. Three
samples were evaluated, a blind control milk (same as
reference) and milk samples spiked with the terpenoid
compounds at concentrations determined at d 0 and at
d 14, along with a reference milk; the results are shown
in Table 6. In comparison to the blind control reference
milk, the milk spiked with compounds at equivalent
concentrations to d 0 SOF milk samples showed an
increased difference score (mean score = 1.7, P-value
= 0.06), whereas the milk spiked with d 14 levels was
highly statistically different (mean score = 2.0, P-value
< 0.01). The panelists described the spiked milk sam-
ples as “stale,” “pine-like,” “dirty,” “cardboard,” “old,”
“unclean,” “musty,” “earthy,” “grassy,” and “metallic.”
Descriptors such as stale, cardboard, and grassy have
historically been used to describe purported SOF milk
samples as well (Shipe et al., 1978; Nicholson, 1993;
Juhlin et al., 2010b); however, in the current study the
source of these objectionable flavors was found to be of
microbial origin, and not lipid oxidation.
Microbial flavor taints can arise from microorganisms
of aqueous or terrestrial sources within the milk sup-
ply chain, for example through environmental contact
or contamination, feed or water sources, silage, the
milking process, comingled milk at the processing site,
and so on (Siegmund and Pöllinger-Zierler, 2006). The
SOF milk has frequently been associated with seasonal
occurrences, which points to a variability in animal
conditions or husbandry practices. A seasonal altera-
tion could likewise result in microbial contamination,
growth, and subsequent negative flavor development in
milk. Additionally, thermoacidophilic bacteria capable
of surviving pasteurization could result in changes in
milk quality during subsequent storage. This would ac-

Table 5. Aroma compounds identified in the neutral/basic fraction of spontaneous oxidized flavor labeled milk1

Average OSME intensity

Identified compound   DB-WAX2   HP-52   Odor descriptor d0 d 14

Propanoic acid 1523 687 Pungent, rancid — 1.8
Pentanal — 732 Pungent 2.2 2.0
Hexanal 1107 777 Oxidized, fat 1.7 1.0
Butyric acid — 816 Cabbage, onion — 1.7
Hexanol 1347 844 Hay, grass 1.7 1.8
2-Heptanone 1195 882 Fatty 1.7 2.3
3-Octanol 1347 981 Mushroom, earthy 2.0 2.0
Benzyl alcohol 1872 1040 Roasted, meaty 2.2 1.8
1-Octanol 1550 1076 Chemical, metal 2.3 1.7
2-Nonanone 1435 1090 Green 2.3 2.0
Nonanal 1347 1112 Fat, cream 2.0 2.3
Endo-borneol 1657 1169 Paper, musty 2.2 2.7
2-Methylisoborneol 1573 1175 Musty 2.3 3.7
Octanoic acid — 1179 Sweat, cheese 2.3 2.5
Naphthalene 1696 1189 Musty — 2.5
α-Terpineol 1678 1192 Musty 2.3 2.3
Decanal 1550 1197 Green, fat 2.3 2.2
E-2-Decenal 1573 1240 Cheese, fat 1.8 2.0
E,E-2,4-decadienal 1642 1252 Paper, cardboard, bean 2.0 2.0
δ-Octalactone — 1270 Peach, coconut 2.2 2.3
Decanol 1696 1272 Fat 2.0 2.7
δ-Nonalactone 2163 1281 Coconut 2.3 2.7
p-Vinyl guaiacol 2280 1344 Malty, mushroom 2.2 2.8
γ-Nonalactone — 1366 Coconut, peach 1.8 2.3
Decanoic acid — 1373 Rancid, fat 2.5 2.2
δ-Decalactone — 1500 Lactone, cream 2.5 2.7
Dodecanoic acid — 2169 Oil, soap — 1.8
1
Identified on 2 column chemistries with corresponding odor descriptors and OSME odor intensity values from d 0 and 14 post-HTST pasteuri-
zation, averaged among 3 panelists.
2
DB-WAX and HP-5: Agilent Technologies, Santa Clara, California.

Journal of Dairy Science Vol. 101 No. 12, 2018

8 POTTS AND PETERSON

Figure 1. Concentrations of lipid oxidation compounds in off-flavored spontaneous oxidized flavor labeled milk, d 0 and 14. Average concen-
trations are reflected in triplicate with respective SD. Average concentrations marked with an asterisk (*) were significantly different between
d 0 and 14 (P < 0.05).

Figure 2. Concentrations of terpenoid compounds present in off-flavored spontaneous oxidized flavor labeled milk, d 0 and 14. Average
concentrations are reflected in triplicate with respective SD. Average concentrations marked with an asterisk (*) were significantly different
between d 0 and 14 (P < 0.05).

Journal of Dairy Science Vol. 101 No. 12, 2018

 OBJECTIONABLE FLAVORS IN BOVINE MILK
Table 6. Sensory degree of difference of a reference milk, a blind
control milk, and milk spiked with 2 levels of terpenoid compounds
reported in milk at d 0 and 14
Milk sample Mean rating P-value
Blind control 0.7 — Colahan-Sederstrom, P. M., and D. G. Peterson. 2005. Inhibition of
Recombinant d 0 1.7 0.06
Recombinant d 14 2.0 <0.01
1 Chem. 53:398–402.
Panelists evaluated samples using a 5-point scale: 0 = no difference, 1
= very slight difference, 2 = slight difference, 3 = moderate difference,
4 = large difference, or 5 = extremely large difference; P-value using a
2-sided Dunnett’s multiple comparisons test.
count for changes in the flavor taint profile over time,
as reported in Figure 2.
CONCLUSIONS
After further investigation of the objectionable fla-
vored milk flagged as SOF, the source was reported
to arise from microbial flavor taints endo-borneol,
2-methylisoborneol, and α-terpineol. Lipid oxidation
compounds, although identified in both the control and
SOF labeled milk samples in this study, were typically
of high quality milk and not shown to reproduce the ob-
jectionable flavor profile. Therefore, the SOF milk was
inaccurately labeled, indicating the challenges of classi-
fying and investigating related flavor problems without
adequate identification of the active compounds. More
work is needed to understand the prevalence of mi-
crobial taints in the milk supply, the effect of thermal
processing on microbial survival, and the source of the
microbial strains involved. Furthermore, characterizing
the flavor profiles from other SOF-labeled milk could
potentially reveal other sources for the noted flavor
defects. Identification of the sources of objectionable
flavors in milk provides critical control points in dairy
production and distribution to maintain high product
quality.
ACKNOWLEDGMENTS
The authors of this study thank the Midwest Dairy
Association (St. Paul, MN) and the Agriculture Utiliza-
tion Research Institute (Crookston, MN) for provided
support and resources for completion of this project.
REFERENCES
Bruhn, J. C., A. A. Frank, and G. S. Goble. 1976. Factors relating to
development of spontaneous oxidized flavor in raw milk. J. Dairy
Sci. 59:828–833.
9
Buttery, R. G., J. G. Turnbaugh, and L. C. Ling. 1988. Contribution of
volatiles to rice aroma. J. Agric. Food Chem. 36:1006–1009.
Clausen, M. R., C. Connolly, L. H. Skibsted, and J. Stagsted. 2010.
Oxidative stability of bovine milk determined by individual vari-
ability in herd irrespective of selenium status. Int. Dairy J. 20:507–
513.
key aroma compound generated during ultrahigh-temperature pro-
cessing of bovine milk via epicatechin addition. J. Agric. Food
Dunkley, W. L., M. Ronning, and J. Voth. 1963. Influence of injection
of cows with copper glycinate on blood and milk copper and oxi-
dized flavor in milk. J. Dairy Sci. 46:1059–1063.
Engel, W., W. Bahr, and P. Schieberle. 1999. Solvent assisted flavour
evaporation-a new and versatile technique for the careful and di-
rect isolation of aroma compounds from complex food matrices.
Eur. Food Res. Technol. 209:237–241.
Granelli, K., P. Barrefors, L. Björck, and L.-A. Appelqvist. 1998. Fur-
ther studies on lipid composition of bovine milk in relation to
spontaneous oxidised flavour. J. Sci. Food Agric. 77:161–171.
Juhlin, J., W. F. Fikse, I.-L. Orde-Ostrom, P. Barrefors, and A. Lun-
den. 2010a. Factors relating to incidence of spontaneous oxidized
flavor and copper in cow’s milk. Acta Agric. Scand. 60:94–103.
Juhlin, J., F. Fiske, A. Lundén, J. Pckova, and S. Agenäs. 2010b. Rela-
tive impact of α-tocopherol, copper and fatty acid composition on
the occurrence of oxidized milk flavour. J. Dairy Res. 77:302–309.
Karagül-Yüceer,, Y., M. Drake, and K. Cadwallader. 2001. Aroma-
active components of nonfat dry milk. J. Agric. Food Chem.
49:2948–2953.
McDaniel, M. R., R. Miranda-Lopez, B. T. Watson, N. J. Micheals,
and L. M. Libbey. 1990. Pinot noir aroma: A sensory/gas chro-
matographic approach. Page 23 in Flavors Off-Flavors '89. G. Ch-
aralambous, ed. Elsevier, Amsterdam, the Netherlands.
Nicholson, J. W. G. 1993. Spontaneous Oxidized Flavour in Cow’s
Milk. International Dairy Federation, Schaerbeek, Belgium.
Plotto, A., C. A. Margaría, K. L. Goodner, R. Goodrich, and E. A.
Baldwin. 2004. Odour and flavour thresholds for key aroma com-
ponents in an orange juice matrix: Terpenes and aldehydes. Fla-
vour Fragr. J. https:​/​/​doi​.org/​10​.1002/​ffj​.1470.
Schrader, K. K., and S. T. Summerfelt. 2010. Distribution of off-flavor
compounds and isolation of geosmin-producing bacteria in a series
of water recirculation systems for rainbow trout culture. N. Am.
J. Aquaculture 72:1–9.
Shipe, W. F., R. Bassette, D. D. Deane, W. L. Dunkley, E. G. Ham-
mond, W. J. Harper, D. H. Kleyn, M. E. Morgan, J. H. Nelson,
and R. A. Scanlan. 1978. Off flavors of milk: Nomenclature, stan-
dards, and bibliography. J. Dairy Sci. 61:855–869. https:​/​/​doi​.org/​
10​.3168/​jds​.S0022​-0302(78)83662​-5.
Siegmund, B., and B. Pöllinger-Zierler. 2006. Odor thresholds of mi-
crobially induced off-flavor compounds. J. Agric. Food Chem.
54:5984–5989.
Teranishi, R., R. G. Buttery, and D. G. Gaudagni. 1974. Odor quality
and chemical structure in fruit and vegetable flavors. Ann. N. Y.
Acad. Sci. 237:209–216.
Timmons, J. S., W. P. Weiss, D. L. Palmquist, and W. J. Harper.
2001. Relationships among dietary roasted soybeans, milk com-
ponents, and spontaneous oxidized flavor of milk. J. Dairy Sci.
84:2440–2449.
Villeneuve, M.-P., Y. Lebeuf, R. Gervais, G. F. Tremblay, J. C. Vuil-
lemard, J. Fortin, and P. Y. Chouinard. 2013. Milk volatile or-
ganic compounds and fatty acid profile in cows fed timothy as hay,
pasture, or silage. J. Dairy Sci. 96:7181–7194. https:​/​/​doi​.org/​10​
.3168/​jds​.2013​-6785.
Journal of Dairy Science Vol. 101 No. 12, 2018